PROGNOSTICS AND HEALTH MANAGEMENT

OF PV PANEL USING MACHINE INTELLIGENCE

18PS380 Dissertation Phase 1

Submitted in partial fulfillment for the requirement of

M.E degree in power system engineering
Anna University

Submitted by
K. SOFFIAH
(Reg. No: 19K013)

Guided by
Dr. P.S. MANOHARAN
Associate Professor
Department of Electrical and Electronics Engineering

THIAGARAJAR COLLEGE OF ENGINEERING

(A Govt. Aided Autonomous Institution affiliated to Anna University)

MADURAI – 625 015

June 2020
i
 THIAGARAJAR COLLEGE OF ENGINEERING
(An Autonomous Institution Affiliated to Anna University)

MADURAI – 625 015

CERTIFICATE

Certified that this is a bonafide record of the 18PS280 Mini Project done by
Ms.K.SOFFIAH ( Reg.No. 19K013 ) of Second Semester M.E (Power System
Engineering) during the year 2020.

Dr. P.S. Manoharan, Dr. N. Kamaraj

Associate Professor Professor & Head,
Department of EEE. Department of EEE
(Guide)

Station: Madurai Date:

Submitted for VIVA-VOCE examination held at Thiagarajar College of

Engineering, Madurai – 625 015, on __________________.

INTERNAL EXAMINER EXTERNAL EXAMINER H.D.E.E

ii
 BONAFIDE CERTIFICATE

This is to certify that the 18PS380 Dissertation Phase 1 Report entitled “Prognostics and

Health Management of PV Panel using Artificial Neural Network”, being submitted by K.

SOFFIAH in partial fulfillment for the requirement of M.E. Degree in Power System

Engineering is a record of bonafide work. The results embodied in this report have not been

submitted to any other university or institute for the award of any degree or diploma.

DR. P.S. MANOHARAN, DR.N. KAMARAJ,

Associate Professor Professor & Head,
Department of EEE. Department of EEE
(Guide)

Station: Madurai Date:

iii
 ACKNOWLEDGEMENT

It is my proud privilege to express my deep sense of gratitude towards all those who gave
me the possibility to complete this thesis.

First and foremost, I thank the Almighty for showering his blessings on me throughout
this work.

I would like to express my sincere thanks and gratitude to Dr. S. MERCY SHALINIE,
Principal, Thiagarajar College of Engineering, for his encouragement to accomplish this project.

I wish to express my deep sense of gratitude to DR. N. KAMARAJ, Professor and Head
of the Department, Electrical and Electronics Engineering, for granting me permission to do this
project and for his motivation, suggestions and trust made on me.

I would like to express my deep sense of gratitude to my guide DR. P.S.

MANOHARAN, Associate Professor, Electrical and Electronics Engineering, for his persistent,
valuable guidance, keen suggestions, innovative ideas and unflinching support at each and every
stage of my work.

I would like to express my special gratitude and thanks to staff members of Electrical and
Electronics department for giving me such attention and time.

Finally, I would like to express my special gratitude towards my parents for their kind co-
operation and encouragement which brought courage in me to complete this project. My thanks
and appreciations also go to my friends in developing the project and people who have willingly
helped me out with their abilities.

The technical information derived from various research papers is gratefully

acknowledged.

iv
 ABSTRACT

The surge in renewable electricity generation using photovoltaic (PV) systems was
accompanied by an increased awareness of the fault conditions developed during the operational
lifetime. Fault detection, diagnostics, and prognostics are such efforts to detect and classify a
fault so the system operational expectations can be managed. Trending of the faults and
prognostics also aid to evaluate expected remaining useful life so that mitigation actions can be
evaluated and implemented. This proposed method involves Artificial Neural Network to
monitor, detect and classify the faults that occurs at PV panel.

v
 TABLE OF CONTENTS
CHAPTER NO. PAGE NO.
ABSTRACT v
TABLE OF CONTENT vi
LIST OF FIGURE viii
LIST OF TABLES ix
1 INTRODUCTION 1
1.1 POSSIBLE FAILURE MODES IN PV 1
SYSTEM
1.2 PV MODULE 1
1.2.1 DEGRADATION 2
1.2.2 HOTSPOT 3
1.2.3 BUBBLES 4
1.2.4 SHADING AND SOILING 4
1.3 CABLING 5
1.3.1 GROUND FAULTS 5
1.3.2 LINE – LINE FAULTS 5
1.3.3 ARC FAULTS 6
2 LITERATURE SURVEY 7
3 PROBLEM DESCRIPTION 17
3.1 PROBLEM IDENTIFIED 17
4 PROPOSED TECHNOLOGY 18
4.1 ARTIFICIAL INTELLIGENCE 18
4.2 ARTIFICIAL NEURAL NETWORK 18
4.3 LEARNING METHODS 19
4.3.1 SUPERVISED LEARNING 19
4.3.2 UNSUPERVISED LEARNING 19
4.3.3 REINFORCEMENT LEARNING 20
5 BLOCK DIAGRAM 22
5.1 SCHEMATIC PHM BLOCK DIAGRAM 22
6 RESEARCH GAP 23
6.1 RESEARCH GAP 23
7 SIMULATION 24
7.1 MODEL SIMULATION TO COLLECT 24
DATA TO FEED ARTIFICIAL NEURAL
NETWORK MODEL
7.2 MPPT ALGORITHM 26
7.3 SAMPLE DATA COLLECTION 27
7.4 V-I AND P-V CHARACTERISTICS 28
8 WORK ESTIMATION 29
8.1 ESTIMATED WORK OF PHASE 1 29
8.2 ESTIMATED WORK OF PHASE 2 29

vi
 9 CONCLUSION 30

REFERENCES 31

LIST OF FIGURES
FIGURE NO. TITLE PAGE NO.
4.1 SIMPLE ARCHITECTURE OF NEURAL NETWORK 18

4.2 SUPERVISED LEARNING 20

4.3 UNSUPERVISED LEARNING 20

4.4 REINFORCEMENT LEARNING 21

5.1 PHM BLOCK DIAGRAM 21

7.1 SIMULATION TO FIND V-I AND P-V CHARACTERISTICS OF THE 24
PANEL FOR GIVEN IRRADIANCE
7.2 9 SOLAR CELLS IN SERIES CONNECTION IN EACH SUBSYSTEM 25
7.3 SUBSYSTEMS CONNECTED IN SERIES WITH 9 SOLAR CELLS 25
EACH TO FORM PANEL
7.4 V-I CHARACTERISTICS AT VARIOUS IRRADIANCE AND 28
TEMPERATURE AT 25 DEGREE CELSIUS

vii
7.5 P-I CHARACTERISTICS AT VARIOUS IRRADIANCE AND 28
TEMPERATURE AT 25 DEGREE CELSIUS

LIST OF TABLES

TABLE NO. TABLES PAGE NO.

7.1 SAMPLE DATA 27

viii
 CHAPTER 1

INTRODUCTION

Photovoltaic (PV) systems have attracted more attention as prominent renewable energy
sources. This phenomenon is driven by several reasons for technology development, sustainable
energy provision and environmental concern.

About the global environmental concern, PV system can play important role since it is
environmentally clean. The great consequences of this achievement is implicated through the
huge installation of building integrated PV systems and large scale arrays of PV modules
installed in rural sites in the near future. As the size of PV array increases or the installation is
remotely accessed either in open-rural sites or roof-top buildings, the quick and efficient
monitoring operation of the whole PV system including the physical condition of cells will be
attracted to be solved.

Every single module must be ensured to operate properly in order to keep the continuous
power supply and to avoid potential damage.

1.1 POSSIBLE FAILURE MODES IN PV SYSTEM

1.2 PV MODULE

Module mismatch causes each module to operate at a suboptimal point on the Current-
Voltage (I-V) curve, reducing the array’s power output. I-V mismatches are grouped, according
to the causes, as permanent or temporary. Permanent mismatch is due to the effects of changes in
one or more parameters of the PV module, such as the value of parallel resistance and/or series
resistance. In addition to the manufacturing tolerance, module degradation, hot spot, and bubbles
mainly cause permanent mismatches.

1
 Temporary mismatches are affected by the temporal changes in the irradiance level
received by PV modules. Such changes include cloud effects, soiling, snow covering, leaf and
bird droppings, and the shading from nearby PV arrays and structures.

1.2.1 DEGRADATION

The degradation and aging of a PV module is a continuous process, but several factors
can influence its dynamics. In particular, environmental factors such as Sulphur, acidic fumes, or
other pollutants can speed up the degradation process. The main degradation types taking place
in PV modules are as follows:

1.2.1.1 DISCOLOURATION

Discolouration is the browning and yellowing of PV cells, mainly caused by the

degradation of the ethylene vinyl acetate (EVA) encapsulant. The main reasons of EVA
degradation are Ultraviolet (UV) rays combined with water under temperatures higher than 50
°C. The changes in the color of the encapsulant material produce a variation of the transmittance
of the light reaching the solar cells and, as a consequence, a reduction of the power generated.

1.2.1.2 DELAMINATION

Delamination is defined as the breakdown of the bonds between material layers that
constitute a module laminate. Delamination interrupts efficient heat dissipation and increases the
possibility of reverse-bias cell heating. The main causes of delamination are the movement of
cells and cell interconnects due to environmental stresses, the expansion and the contraction of
moisture and air that are trapped inside the layers of a PV module, the bond failure due to the
combination of moisture and UV radiation, the cell overheating, and the consequent outgassing
of the encapsulant. In addition, physical aging processes related to the application of high
temperatures could provoke delamination of a PV module.

2
1.2.1.3 CRACKING

Cracking is a common problem encountered in PV modules. It may develop in different

stages of the module lifetime; however, it occurs in most of the cases during installation,
maintenance, and especially during the transportation of modules to their sites. In addition,
cracking is affected by the high-temperature thermal stresses of a cell and thermal cycling
induced thermomechanical stresses, mechanical loads due to wind (pressure and vibrations) and
snow (pressure).

1.2.1.4 CORROSION

Corrosion attacks the metallic connection of PV cells causing a loss of performance by

increasing leakage currents. The moisture that enters the module through the laminate edges
mainly causes corrosion. The corrosion of the conductive parts of the cells and the
interconnections through the encapsulant is responsible for the deterioration of the PV module,
which results in the increase of the series resistance and the decrease of the parallel resistance of
the PV electrical model.

1.2.2 HOT SPOT

The short-circuit current and the open circuit voltage are imposed by the PV cell showing
the lowest electrical performance respectively in series and parallel montage. In short circuit
conditions, when a PV cell is defective, its voltage is reversed and becomes equal and opposite to
the voltage of the other cells in series. This defective cell becomes both a load for other cells and
a place of a relatively high thermal dissipation constituting thus a hot spot. A hot spot is an area
of a PV module that has a very high temperature that could damage a cell or any other element of
the module. It occurs in a PV module when the current capability of a particular cell or cells is
lower than the operating current of the cell string. Over time, hot spots will permanently degrade
the PV panels and decrease the overall performance of the PV plant.

3
1.2.3 BUBBLES

The bubbles are mainly due to the chemical reactions that emit gases trapped in the PV
module. They form an air chamber in which the gas temperature is lower than in the adjacent
cells. However, the air chamber worsens the heat dissipation capability of the nearby cell so that
the latter overheats and therefore exhibits a temperature that is higher than in the adjacent cells .
Moreover, when bubbles appear on the front side, a reduction of the radiation reaching the PV
cell occurs, thus creating a decoupling of light and increasing the reflection. Furthermore,
bubbles can break, and can damage the back sealing surface that provokes humidity ingress.

1.2.4 SHADING AND SOILING

Shading, the total or partial blockage of sunlight from a PV module surface, can bring
serious concern in PV arrays. This blockage can be caused by a number of different reasons, like
shade from the building itself, light posts, trees, dirt, snow and other light blocking obstacles.
Shading causes large performance drops and can even damage modules if not properly
controlled. Module soiling is the build-up of dirt on the surface of a PV module. Researchers
have found that the effects of soiling are relatively small (2.3% loss of power) for directly
incident light but become more significant for larger angles: an 8.1% loss was observed in a
soiled module when light is incident from an angle of 56◦. An experimental investigation on the
reduction of PV output efficiency showed that the reduction of efficiency reached up to 11.6%
when the dust deposition density was fixed at about 8 g/m2. In addition, a single dust storm can
reduce the output power by 20% and a reduction of 50% could be experienced if no cleaning is
performed on modules for long time that exceeds six months. The local soil and environmental
conditions are key factors for severity impact.

4
1.3.. CABLING

Cables are vital parts of a PV array. Similar to the rest of the PV system, cables are
subjected to thermal, mechanical and external loads. Though the selection of cables is an
important procedure, cable terminations and cables management thereafter can influence how the
entire PV system will function. Three major catastrophic failure modes are common in the
cabling of PV systems: ground faults, line-line faults, and arc faults.

1.3.1 GROUND FAULTS

A ground failure mode occurs when the circuit develops an unintentional path to ground.
This results in lowered output voltage and power, and can be fatal if the leakage currents are
running through a person. If a ground fault remains undetected, it may generate a DC arc within
the fault and cause a fire hazard. Previous research investigated the potential reasons that can
lead to ground faults, and classified them into four categories:

 Cable insulation damage during the installation, due to aging, impact damage, water leakage,
and corrosion;
 Ground fault within the PV modules (e.g., degraded sealant and water ingress);
 Insulation damage of cables due to chewing done by rodents and termites; and
 Accidental short circuit inside the PV source circuit combiners, often at the time of
maintenance.

1.3.2 LINE-LINE FAULTS

A line-to-line failure mode in a PV system is defined as an unintentional connection

between two points in a PV panel through a low resistance path. However, if one of the points is
on the Equipment Grounding Conductor (EGC), the line-to-line fault is considered as a ground
fault. A line-to-line fault may occur between two points on the same string or between two
adjacent strings. The magnitude of the line-to-line fault current depends on the potential

5
difference between the points before the fault occurs. The higher the potential difference, the
higher the back feed current results, and the chance of tripping increases. Several studies
summarize the reasons behind line-to-line faults in PV arrays as follows:
 Insulation failure of cables, i.e. UV degradation, animal chewing through cable insulation;
 Incidental short circuit between current carrying conductors, i.e. a nail driven through
unprotected wirings; and
 Line-line faults within the DC junction box, which are caused by mechanical damage, water
ingress or corrosion.

1.3.3. ARC FAULTS

Arc failure mode establishes a current path in the air, and this current path might be
established due to any discontinuity in the current carrying conductors or insulation breakdown
in adjacent current carrying conductors. Any type of arc fault is harmful for the PV system, and
may introduce fire that may result in insulation burn-out and fire hazards in presence of any
flammable substances in the vicinity of the PV plant. National Electrical Code® (NEC)-2011
requires a series arc-fault protection device in a PV system if the DC operating voltage is equal
to or higher than 80V. These devices are called as arc-fault circuit interrupters (AFCIs). The
causes of arc faults depend on their types, whether they are series or parallel. Series arc fault
reasons include degradation in solder joints, wiring or connections inside the junction box,
loosening of screws, and increased operating temperature that may result in thermal stress,
leading to accelerated aging or complete disconnection. In addition to series arc-fault reasons,
parallel arc faults can result from insulation damage due to mechanical damage, aging, or
wildlife.

6
 CHAPTER 2

LITERATURE SURVEY

2.1 MODELING AND FAULT DIAGNOSTIC OF A PHOTOVOLTAIC SYSTEM

Authors:

Kuei-Hsiang Chaoa, Sheng-Han Ho, Meng-Hui Wang

Journal:

Electric Power Systems Research 78 (2012) 97–105

Inference:

In this paper, an accurate circuit-based PV module was established by the PSIM software
package, which combined a 3kW PV arrays system as well as gathered the tested data for fault
diagnosis. According to the compared results, the proposed PSIM based PV module possesses a
higher accuracy in electrical parameters than the conventional mathematic model. Furthermore,
the simulated results also show that the proposed fault diagnosis method can easily recognize the
main fault category and indicate the possibilities of others.

Drawback:

Further work is needed to identify the fault location.

7
2.2 A NOVEL FAULT DIAGNOSIS TECHNIQUE FOR PHOTOVOLTAIC SYSTEMS
BASED ON ARTIFICIAL NEURAL NETWORKS

Authors:

W. Chine , A. Mellit , V. Lughi

Journal:

Renewable Energy 90 (2016) 501-512.

Inference:

The paper presents a new Artificial Neural Network (ANN) based approach for the
identification of eight types of fault occurring in a PV array. Moreover, this work also shows the
implementation of the proposed fault diagnosis technique into a Field Programmable Gate Array
(FPGA) showing its effectiveness in real applications. Different attributes (such as current,
voltage and number of peaks)
of the simulated and the measured current-voltage (I-V) characteristics of a number of PV strings
have firstly been compared. The development of two different algorithms allows the isolation
and the identification of faults that have and have not the same combination of attributes. The
obtained results confirm the ability of the technique to correctly localize and identify the
different type of faults.

Drawback:

The model failed to take environmental condition into account.

8
2.3 AUTOMATIC SUPERVISION AND FAULT DETECTION OF PV SYSTEMS
BASED ON POWER LOSSES ANALYSIS

Authors:

A. Chouder , S. Silvestre

Journal:

Energy Conversion and Management 51 (2010) 1929–1937

Inference:

A new automatic supervision and fault detection procedure for PV systems, based on the
power losses analyse has been presented. The tool, developed in Matlab, includes parameter
extraction techniques to calculate main PV system parameters from monitoring data in real
conditions of work, taking into account the environmental irradiance and module temperature
evolution in the simulation model. The main idea of the proposed system diagnosis and fault
detection of a PV system is based in the continuous check of the measured capture losses.

Drawback:

The proposed system fails to update parameters regarding changing environmental

condition.

9
2.4 INTELLIGENT FAULT DETECTION AND DIAGNOSTICS IN
SOLAR PLANTS

Authors:

Andon Coleman, Janusz Zalewski

Journal:

The 6th IEEE International Conference on Intelligent Data Acquisition and Advanced
Computing Systems: Technology and Applications
15-17 September 2011, Prague, Czech Republic.

Inference:

The contribution of this work is the development of a fault detection and diagnostics
system based on Bayesian belief networks. The current implementation establishes a framework
for validating sensor readings and mapping them to a Bayesian belief network using the Netica
API – it lays the groundwork for future improvement.

Drawback:

10
 Need concentration on more accurately distributing the probabilities of failure and
reducing the complexity associated with increasing the number of modeled measurements.

2.5 SOLAR ARRAY FAULT DETECTION USING NEURAL NETWORKS

Authors:

Sunil Rao, Andreas Spanias and Cihan Tepedelenlioglu

Journal:

IEEE Interrnational Conference on industrial Cyber Physical Systems,2019.

Inference:

This paper, describe a Cyber-Physical system approach to fault detection in Photovoltaic

(PV) arrays. More specifically, this paper explore customized neural network algorithms for fault
detection from monitoring devices that sense data and actuate at each individual panel. This
system develop a framework for the use of feed forward neural networks for fault detection and
identification.

Drawback:

11
 The proposed system fails to update parameters regarding changing environmental
condition.

2.6 PHOTOVOLTAIC PROGNOSTICS AND HEATH MANAGEMENT USING

LEARNING ALGORITHMS

Author:

Daniel Riley and Jay Johnson

Journal:

38th IEEE Photovoltaic Specialists conference,2012.

Inference:

This paper proposed a novel model-based prognostics and health management (PHM)
system to monitor the health of a photovoltaic (PV) system, measure degradation, and indicate
maintenance schedules. The proposed method uses an artificial neural network (ANN) which
eliminates the need for a priori information by teaching the algorithm "good" performance
behavior based on the initial performance of the array. The PHM algorithm was trained using
two months of AC power production. The model then predicted the output power of the system
using irradiance, wind, and temperature data. Based on the deviation in measured AC power
from the AC power predicted by the trained ANN model, system outages and other faults
causing a reduction in power were detected.

12
2.7 CONTROLLING OF ARTIFICIAL NEURAL NETWORK FOR FAULT
DIAGNOSIS OF PHOTOVOLTAIC ARRAY

Author:

Syafaruddin,E. Karatepe, T. Hiyama

Journal:

16th International Conference on Intelligent System Applications to Power

Systems,2011.

Inference:

This paper has proposed fault diagnosis of PV array, especially for short-circuit condition
using three layered feed forward neural network. A single artificial neural network (ANN) model
is less suitable since the error might be high, while in the diagnosis solving problems the high
accuracy is the most important issue. Therefore, several ANN model are developed based on the
estimation task, then control rule is provided to drive to which ANN can respond to the fault
location.

13
Drawback:

In the current form, the proposed method is able to identify the short-circuit location of
PV modules in one string independently. In order to extent the task of the proposed method, for
instance to find the multiple location of fault between strings, the number of data training should
be increased and the control rule should be updated.

2.8 MONITORING SOLAR PANELS USING MACHINE LEARNING

TECHNIQUES

Author:

Cristian-Győző Haba

Journal:

2019 8th International Conference on Modern Power Systems (MPS).

Inference:

This paper proposed a method of monitoring photovoltaic panels for the identification of
possible degradation of operation based on machine learning (ML) techniques. The development
of the model and its validation is based on data collected both from the panels' photovoltaic
control system and weather data collected from Internet.

Drawback:

14
 The possible faults that has been taken in to account includes only PV panel degradation
fault.

2.9 ENVIRONMENTAL DEGRADATION OF THE OPTICAL SURFACE OF PV

MODULES AND SOLAR MIRRORS BY SOILING AND HIGH RH AND MITIGATION
METHODS FOR MINIMIZING ENERGY YIELD LOSSES

Authors:

Malay K. Mazumder, Mark N. Horenstein, Calvin Heiling, Jeremy W. Stark, Arash

Sayyah, Julius Yellowhair, and Atri Raychowdhury.

Journal:
2015 IEEE 42nd Photovoltaic Specialist Conference (PVSC)

Inference:

This paper provide a brief review of the application of an emerging technology of

transparent electrodynamic screen (EDS) for removing dust, as frequently as needed, from the
solar collectors without water. Power output efficiency is maintained greater than 90% compared
to that of the panel under clean conditions. Dust removal efficiency (DRE) is more than 90%

15
with test dust samples obtained from different arid zones and energy consumption for EDS
operation is less than 0.03 Wh/m2/cleaning cycle. The method is water-free and
provides easy retrofitting onto existing panels and has a high potential for a cost-effective large-
scale roll-to-roll production, commercial application, and a significant reduction of operation and
maintenance costs.

2.10 A NEW FAST DETECTION MODULE FOR SHORT-CIRCUIT CURRENT

DETECTION IN PV GRID SYSTEM

Authors:

Krishnakant Kasar, Prof. P.C.Tapre.

Journal:

Proceedings of the Second International Conference on Inventive Systems and Control

(ICISC 2018).

Inference:

In this paper, a new fast detection module for short-circuit current detection has been
discussed based on the rate of rise of current together with the current magnitude in a PV solar
system based DG. There can be two control operations, as per the applicable grid code in that
region, one of which is the disconnection of the PV inverter before the current exceeds the rated
output current of the inverter. The second one discusses about transforming the PV inverter into
a dynamic reactive power compensator STATCOM and provide grid support functions. A new

16
concept of utilizing PV solar farms as STATCOM (PV-STATCOM) both during nighttime and
daytime for different grid support functions was introduced in this paper.

CHAPTER 3

PROBLEM DESCRIPTION

3.1 PROBLEM IDENTIFIED

• Reliability is a critical factor for a PV system.

• Faults at PV array leads to reduced and inefficient functioning of the PV array.
• Inefficient functioning significantly lowers the power output.
• In solar power plants, monitoring health and occurrence of fault at PV panel manually is
quiet difficult.
• The mean time to repair is high.
• Thus an Artificial Neural Network can be used to detect and identify the occurred fault
efficiently.

17
 CHAPTER 4

PROPOSED TECHNOLOGY

4.1 ARTIFICIAL INTELLIGENCE

Artificial intelligence is composed of systems that allow computers to imitate human

cognitive processes or perform tasks that used to be done by humans.

4.2 ARTIFICIAL NEURAL NETWORK

An artificial neural network (ANN) is the piece of a computing system designed to
simulate the way the human brain analyzes and processes information. It is the foundation of
artificial intelligence (AI) and solves problems that would prove impossible or difficult by
human or statistical standards.

Artificial neural network is an efficient information processing system which resembles

characteristics of biological neural network. ANN consists of consists of highly interconnected
processing element called nodes or neurons and operates in parallel. Each neuron is connected
with other through connection link. Each connection link contains weights which has the

18
information regarding the input signal. This information is used by the neuron to solve the
particular problem.

Figure 4.1 Simple Architecture of Neural Network

y= x1w1+x2w2  Equation 1

Where,

• X1,X2 are input neurons

• Y is output neuron
• x1,x2 are activation signal
• w1,w2 are weights
• y is output signal.

4.4 LEARNING METHODS

4.4.1 SUPERVISED LEARNING

Supervised learning algorithms are trained using labeled examples, such as an input where the
desired output is known. For example, a piece of equipment could have data points labeled either
“F” (failed) or “R” (runs). The learning algorithm receives a set of inputs along with the

19
corresponding correct outputs, and the algorithm learns by comparing its actual output with
correct outputs to find errors. It then modifies the model accordingly. Through methods like
classification, regression, prediction and gradient boosting, supervised learning uses patterns to
predict the values of the label on additional unlabeled data. Supervised learning is commonly
used in applications where historical data predicts likely future events. For example, it can
anticipate when credit card transactions are likely to be fraudulent or which insurance customer
is likely to file a claim.

Figure: 4.2 Supervised Learning

4.4.2 UNSUPERVISED LEARNING

Unsupervised learning is used against data that has no historical labels. The system is not told the
"right answer." The algorithm must figure out what is being shown. The goal is to explore the
data and find some structure within. Unsupervised learning works well on transactional data. For
example, it can identify segments of customers with similar attributes who can then be treated
similarly in marketing campaigns. Or it can find the main attributes that separate customer
segments from each other. Popular techniques include self-organizing maps, nearest-neighbor
mapping, k-means clustering and singular value decomposition. These algorithms are also used
to segment text topics, recommend items and identify data outliers.

20
 Figure: 4.3 Unsupervised Learning

4.4.3 REINFORCEMENT LEARNING

Reinforcement learning is often used for robotics, gaming and navigation. With reinforcement
learning, the algorithm discovers through trial and error which actions yield the greatest rewards.
This type of learning has three primary components: the agent (the learner or decision maker),
the environment (everything the agent interacts with) and actions (what the agent can do). The
objective is for the agent to choose actions that maximize the expected reward over a given
amount of time. The agent will reach the goal much faster by following a good policy. So the
goal in reinforcement learning is to learn the best policy.

Figure : 4.4 Reinforcement Learning

21
 CHAPTER 5

BLOCK DIAGRAM

5.1 SCHEMATIC PHM BLOCK DIAGRAM

As in Figure 5.1, the PV PHM system compares the measured metric of interest to a
prediction of the metric from a model. The benefit of using machine intelligence based model is
that the user does not need to know any specifics about the PV system components. Machine
intelligence simply identifies the relationships between the PV system's environment (input) and
its power production. Thereby, comparing the produced power and the expected power,
occurrence of fault can be identified.

22
 Figure 5.1 PHM Block Diagram

CHAPTER 6

RESEARCH GAP

6.1 RESEARCH GAP

• Artificial Neural Network has been used to detect fault yet not modeled to figure the type
of fault .

• Thus a multilayer artificial neural network with relevant conditions can identify the type
of fault.

23
 CHAPTER 7

SIMULATION

7.1 MODEL SIMULATION TO COLLECT DATA TO FEED ARTIFICIAL NEURAL

NETWORK MODEL

A solar panel is constructed using 27 Solar Cells which are grouped into subsystems
with 9 solar cells each. The temperature dependence of each solar cell is fixed at 25 degree
Celsius. The irradiance is given as the input to the subsystem of solar cells using a constant
block. And the resulted output current and voltage for the given irradiance is sensed using
current and the voltage sensor respectively and is display through the display block. And V- I
and P-V characteristics are plotted for the given irradiance.

24
 The MATLAB function block is used to find the maximum voltage and current at
maximum power point.

Figure 7.1 Simulation To Find V-I and P-V Characteristics of the Panel for Given Irradiance

25
 Figure 7.2 9 Solar Cells in Series Connection in Each Subsystem

Figure 7.3 Subsystems Connected in Series with 9 Solar Cells Each to Form Panel

26
7.2 MPPT PROGRAM

function [max_P,max_V,max_I] = max_val(P,V,I)

persistent Ppre Vpre Ipre
if isempty(Ppre)
Ppre =0;
Vpre = 0;
Ipre=0;
end

if(Ppre > P)
max_P = Ppre;
max_V = Vpre;
max_I=Ipre;
else
max_P=P;
max_V=V;
max_I=I;
end
Ppre = max_P;
Vpre = max_V;
Ipre = max_I;

27
7.3 SAMPLE DATA COLLECTION

So far sample data collected are given in the Table 7.1.

Table 7.1 Sample Data

28
7.4 V-I AND P-V OUTPUT CHARACTERISTICS

Figure 7.4 V-I Characteristics at various Irradiance and Temperature at 25 degree Celsius

Figure 7.5 P-V Characteristics at various Irradiance and Temperature at 25 degree Celsius

29
 CHAPTER 8

WORK ESTIMATION

8.1 ESTIMATED WORK OF PHASE 1

In phase one of the proposed project, it is estimated to collect the data regarding the panel
specification, power generated, V-I characteristics and the corresponding environmental
condition and implementing the proposed work using Artificial Neural Network in MATLAB.

8.2 ESTIMATED WORK OF PHASE 2

In phase two of the project, the information gathered from the phase one is incorporated
further using machine intelligence technology by selecting appropriate machine learning method
to detect and diagnose the occurrence of fault.

30
 CHAPTER 9

CONCLUSION

This system keeps continuous monitoring of solar panels ,regular inspection of the panel
become easy and efficient. With the help of this work it is possible to detect any fault
occurred within the panel as the generated power may show some inconsistency in data of
solar power plant. It provides unmanned fault identification system using Machine
intelligence to reduce the work of humans. Thereby, it reduces the mean time to
repair(MTTR).

31
 REFERENCE

• Mohammad AbdulHadi, Member, IEEE, Abdulrahman M. Al-Ibrahim, and Gurvinder S.

Virk, “ Neuro-Fuzzy-Based Solar Cell Model”, 38th IEEE Photovoltaic Specialists
conference,2012.

• Kuei-Hsiang Chaoa,Sheng-Han Hob, Meng-Hui Wang, “ Modeling and fault diagnosis

of a photovoltaic system”, Proceedings of the 6th IEEE International Conference on
Intelligent Data Acquisition and Advanced Computing Systems,2011.

• Sandy Rodrigues,Helena Geirinhas Ramos, and F. Morgado-Dias, “ Machine

Learning in PV Fault Detection, Diagnostics and Prognostics: A Review”, IEEE 44th
Photovoltaic Specialist Conference (PVSC),2017.

• Daniel Riley and Jay Johnson, “Photovoltaic Prognostics and Heath Management using
Learning Algorithms”, 38th IEEE Photovoltaic Specialists conference,2012.

• Sunil Rao, Andreas Spanias and Cihan Tepedelenlioglu , “Solar Array Fault Detection
using Neural Networks ” , IEEE Interrnational Conference on industrial Cyber Physical
Systems,2019.

• A.Chouder S.Silvestre , “Automatic supervision and fault detection of PV systems based

on power losses analysis”, Energy Conversion and ManagementVolume 51, Issue 10,
October 2010

• Syafaruddin,E. Karatepe, T. Hiyama, “Controlling of Artificial Neural Network for

Fault Diagnosis of Photovoltaic Array”, 16th International Conference on Intelligent
System Applications to Power Systems,2011.

• Mohamed Deriche, Mohammed Wasim, Wael Suliman, “An IoT based sensing system
for remote monitoring of PV system”, 16th International Multi-Conference on Systems,
Signals & Devices (SSD) ,2019.

32